cs2no_wav2vec2-large-xls-r-300m-czech-colab
Browse files
README.md
ADDED
|
@@ -0,0 +1,92 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
base_model: facebook/wav2vec2-lv-60-espeak-cv-ft
|
| 4 |
+
tags:
|
| 5 |
+
- generated_from_trainer
|
| 6 |
+
datasets:
|
| 7 |
+
- nb_samtale
|
| 8 |
+
metrics:
|
| 9 |
+
- wer
|
| 10 |
+
model-index:
|
| 11 |
+
- name: cs2no_wav2vec2-large-xls-r-300m-czech-colab
|
| 12 |
+
results:
|
| 13 |
+
- task:
|
| 14 |
+
name: Automatic Speech Recognition
|
| 15 |
+
type: automatic-speech-recognition
|
| 16 |
+
dataset:
|
| 17 |
+
name: nb_samtale
|
| 18 |
+
type: nb_samtale
|
| 19 |
+
config: annotations
|
| 20 |
+
split: test
|
| 21 |
+
args: annotations
|
| 22 |
+
metrics:
|
| 23 |
+
- name: Wer
|
| 24 |
+
type: wer
|
| 25 |
+
value: 0.8457142857142858
|
| 26 |
+
---
|
| 27 |
+
|
| 28 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 29 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 30 |
+
|
| 31 |
+
# cs2no_wav2vec2-large-xls-r-300m-czech-colab
|
| 32 |
+
|
| 33 |
+
This model is a fine-tuned version of [facebook/wav2vec2-lv-60-espeak-cv-ft](https://huggingface.co/facebook/wav2vec2-lv-60-espeak-cv-ft) on the nb_samtale dataset.
|
| 34 |
+
It achieves the following results on the evaluation set:
|
| 35 |
+
- Loss: 396.8153
|
| 36 |
+
- Wer: 0.8457
|
| 37 |
+
|
| 38 |
+
## Model description
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Intended uses & limitations
|
| 43 |
+
|
| 44 |
+
More information needed
|
| 45 |
+
|
| 46 |
+
## Training and evaluation data
|
| 47 |
+
|
| 48 |
+
More information needed
|
| 49 |
+
|
| 50 |
+
## Training procedure
|
| 51 |
+
|
| 52 |
+
### Training hyperparameters
|
| 53 |
+
|
| 54 |
+
The following hyperparameters were used during training:
|
| 55 |
+
- learning_rate: 0.0003
|
| 56 |
+
- train_batch_size: 8
|
| 57 |
+
- eval_batch_size: 8
|
| 58 |
+
- seed: 42
|
| 59 |
+
- gradient_accumulation_steps: 2
|
| 60 |
+
- total_train_batch_size: 16
|
| 61 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 62 |
+
- lr_scheduler_type: linear
|
| 63 |
+
- lr_scheduler_warmup_steps: 500
|
| 64 |
+
- num_epochs: 50
|
| 65 |
+
- mixed_precision_training: Native AMP
|
| 66 |
+
|
| 67 |
+
### Training results
|
| 68 |
+
|
| 69 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
| 70 |
+
|:-------------:|:-----:|:----:|:---------------:|:------:|
|
| 71 |
+
| 3026.3663 | 3.51 | 100 | 472.1026 | 0.9873 |
|
| 72 |
+
| 336.2439 | 7.02 | 200 | 239.3806 | 0.9987 |
|
| 73 |
+
| 208.6184 | 10.53 | 300 | 206.7293 | 0.9917 |
|
| 74 |
+
| 182.6556 | 14.04 | 400 | 221.5585 | 0.8908 |
|
| 75 |
+
| 174.3151 | 17.54 | 500 | 262.3953 | 0.8921 |
|
| 76 |
+
| 140.57 | 21.05 | 600 | 225.9887 | 0.8330 |
|
| 77 |
+
| 114.5967 | 24.56 | 700 | 275.7823 | 0.8495 |
|
| 78 |
+
| 91.2748 | 28.07 | 800 | 314.0284 | 0.8610 |
|
| 79 |
+
| 80.0496 | 31.58 | 900 | 314.4608 | 0.8552 |
|
| 80 |
+
| 66.7338 | 35.09 | 1000 | 326.7965 | 0.8527 |
|
| 81 |
+
| 56.921 | 38.6 | 1100 | 373.0237 | 0.8425 |
|
| 82 |
+
| 50.7125 | 42.11 | 1200 | 374.9553 | 0.8527 |
|
| 83 |
+
| 47.4235 | 45.61 | 1300 | 404.8124 | 0.8489 |
|
| 84 |
+
| 45.1623 | 49.12 | 1400 | 396.8153 | 0.8457 |
|
| 85 |
+
|
| 86 |
+
|
| 87 |
+
### Framework versions
|
| 88 |
+
|
| 89 |
+
- Transformers 4.35.2
|
| 90 |
+
- Pytorch 2.1.0+cu118
|
| 91 |
+
- Datasets 2.15.0
|
| 92 |
+
- Tokenizers 0.15.0
|